SEC16A is a RAB10 effector required for insulin-stimulated GLUT4 trafficking in adipocytes

Insulin regulates whole-body glucose homeostasis in part by stimulating the uptake of glucose into muscle and adipose tissues (Leto and Saltiel, 2012). Insulin stimulates glucose uptake by acutely controlling the amount of glucose transporter 4 (GLUT4) in the plasma membrane (PM; Huang and Czech, 2007; Stöckli et al., 2011). In unstimulated fat and muscle cells, GLUT4 is actively sequestered intracellularly through rapid endocytosis and slowed recycling back to the PM. Upon insulin stimulation, GLUT4 exocytosis is accelerated and internalization is slowed, resulting in a steady-state increase of GLUT4 in the PM and a corresponding increase in glucose uptake (Martin et al., 2000a, 2006; Karylowski et al., 2004). Insulin-stimulated GLUT4 translocation to the PM of muscle and fat cells is blunted in insulin resistance and type 2 diabetes, contributing to systemic dysregulation of glucose metabolism (Abel et al., 2001; Chen et al., 2011).

GLUT4 is ferried to the PM by GLUT4-specialized vesicles (GSVs) that form from the perinuclear recycling endosome/TGN compartment in adipocytes (Leto and Saltiel, 2012). In basal adipocytes, delivery of GSVs to the PM is inefficient, and the GSVs ultimately refuse with endosomes, excluding GLUT4 from the cell surface (Karylowski et al., 2004). Activation of the insulin receptor triggers a series of molecular events that mediate the efficient formation, delivery, docking, and fusion of GSVs to the PM (Gonzalez and McGraw, 2006; Bai et al., 2007; Xiong et al., 2010), resulting in increased GLUT4 at the cell surface. In both basal and insulin-stimulated conditions, PM GLUT4 is in equilibrium with intracellular GLUT4, and therefore, the balance of endocytosis to recycling actively maintains the amount of GLUT4 on the cell surface.

An essential aspect of insulin control of GLUT4 trafficking is inhibition of AS160/TBC1D4, a Rab GTPase-activating protein (GAP; Sano et al., 2003; Eguez et al., 2005; Larance et al., 2005). Knockdown of AS160 results in a fourfold increase of GLUT4 in the PM of basal adipocytes (Eguez et al., 2005). RAB10 is the relevant AS160 target RAB protein in adipocytes that mediates this effect. Knockdown of RAB10 in cultured adipocytes cells or knockout in primary adipocytes results in an ∼50% inhibition of GLUT4 translocation to the PM in response to insulin, and overexpression of RAB10 promotes GLUT4 translocation in a dose-dependent manner, demonstrating RAB10-dependent and -independent insulin regulation of GLUT4 translocation (Sano et al., 2007; Sadacca et al., 2013; Vazirani et al., 2016). In adipocytes, RAB10 has no role in the trafficking of noninsulin responsive cargo, such as the transferrin receptor (TR; Sano et al., 2007; Chen et al., 2012). RAB10 functions at a step before the fusion of GSVs with the PM. Total internal reflection fluorescence (TIRF) microscopy studies have demonstrated that inhibition of RAB10 function blocks insulin-stimulated accumulation of GLUT4-containing vesicles near the PM without inhibiting GLUT4 vesicle fusion with the PM (Bai et al., 2007; Sano et al., 2007; Sadacca et al., 2013).

Identification of the RAB10 effectors involved in GLUT4 trafficking will provide novel insights into the adipocyte response to insulin. To this end, we used affinity chromatography to identify SEC16A as a novel RAB10 effector. We find that SEC16A and GLUT4 are localized to the perinuclear region of the cell with SEC16A concentrated around the perinuclear recycling endosome/TGN compartment containing GLUT4. SEC16A and RAB10 function in insulin-stimulated mobilization of GLUT4 from this compartment, consistent with the RAB10-SEC16A module promoting the formation of GSVs that ferry GLUT4 to the PM. The role of SEC16A in GLUT4 trafficking is independent of its previously characterized role in ER exit site function, because knockdown of other components of the ER exit site machinery, including SEC13, LRRK2, SEC23B, and SEC31, does not affect GLUT4 trafficking. However, knockdown of SEC23A, a component of COPII vesicles, blunted insulin-stimulated GLUT4 translocation but did not affect transport from the ER. These data suggest that a subcomplex of COPII coat proteins are involved in the regulated formation of GSVs that traffic GLUT4 to the PM.

To further explore the role of RAB10 in GLUT4 trafficking, we identified RAB10-GTP–binding proteins by passing lysates from insulin-stimulated 3T3-L1 adipocytes over a column of C-terminal His-tagged RAB10 loaded with GTP-γ-S. After extensive washing, RAB10 and its bound proteins were eluted from the column. Proteins in the elution fraction were identified by mass spectrometry. We focused on six eluted proteins that might be expected to have roles in membrane trafficking: IQGAP1 (IQ motif–containing GAP), ARHGEF7 (RHO GEF 7), GIT1 (an ARF GAP), PFN1 (profilin), ANXA2 (annexin 2), and SEC16A (component of ER exit sites; Table 1). We determined the effects of transiently knocking-down these proteins on insulin response. A 50% or greater knockdown, measured by quantitative RT-PCR, was achieved for each protein (Fig. S1). Only knockdown of SEC16A caused a significant blunting of GLUT4 translocation (Fig. 1, A and B). The effect of SEC16A knockdown was rescued by reexpression of an SEC16A resistant to the siRNA, confirming the effect was caused by depletion of SEC16A (Fig. 1 C).

We determined an ∼50% knockdown of SEC16A based on the cell population mean in the reduction of SEC16A mRNA, whereas we analyzed the effect of the knockdown on GLUT4 translocation at the single-cell level. To assess the relationship between the amount of SEC16A depletion and GLUT4 translocation at the cellular level, we correlated the relative expression of SEC16A in individual cells, measured by quantitative indirect immunofluorescence, to the magnitude of insulin-stimulated GLUT4 translocation. SEC16A siRNA-treated cells were binned into quartiles based on residual SEC16A expression, revealing a graded effect of SEC16A depletion that correlated with an increased inhibition of GLUT4 translocation (Fig. 1, D and E). Representative cells of the lowest and highest quartile expression of SEC16A illustrate the decreased GLUT4 translocation in cells most deplete of SEC16A (Fig. 1 F).

Although SEC16A knockdown had a significant effect on the amount of GLUT4 in the PM of insulin-stimulated adipocytes, SEC16A knockdown did not grossly alter the morphology of the intracellular GLUT4-containing compartments revealed by the GFP fluorescence of the HA-GLUT4-GFP (Fig. 1 G). As is the case in control adipocytes, GLUT4 was distributed between perinuclear compartments and dispersed punctate structures in both basal and insulin-stimulated SEC16A knockdown adipocytes. The perinuclear accumulation of GLUT4 is poorly defined, but previous studies suggest it is an element of the recycling endosomal/TGN compartment (Martin et al., 2000b; Zeigerer et al., 2002; Shi and Kandror, 2005; Huang et al., 2013). Past microscopy studies have used syntaxin 6 as a marker of this compartment, and functional studies have shown a requirement for syntaxin-6 in GLUT4 trafficking to the perinuclear site (Perera et al., 2003; Foley and Klip, 2014). In either basal or insulin-stimulated adipocytes, SEC16A knockdown did not affect targeting of GLUT4 to this syntaxin 6–positive perinuclear compartment, confirming that SEC16A affects the translocation of GLUT4 without grossly altering the intracellular itinerary of GLUT4 (Fig. 1 G).

SEC16A knockdown did not affect insulin-stimulated AKT and AS160 phosphorylation, two components upstream of RAB10 in the insulin signal transduction cascade, demonstrating that the effect of SEC16A knockdown is not secondary to a blunting of insulin signaling (Fig. 2 A).

Knockdown of RAB10 in adipocytes does not affect TR trafficking (Sano et al., 2007; Sadacca et al., 2013). As expected of a RAB10 effector, SEC16A depletion did not alter the amount of TR in the PM of basal or insulin-stimulated adipocytes when analyzed as a population mean or by quartile of SEC16A remaining in the knockdown cells (Fig. 2, B and C). Thus, the effect of SEC16A depletion on GLUT4 trafficking is not secondary to a perturbation in general membrane trafficking.

SEC16A has a role in the export of proteins from the ER (Hughes et al., 2009; Miller and Barlowe, 2010). SEC16A acts to nucleate the formation of COPII vesicles (Watson et al., 2006; Bhattacharyya and Glick, 2007; Iinuma et al., 2007; Whittle and Schwartz, 2010; Cho et al., 2014; Sprangers and Rabouille, 2015).

To investigate if the defect in GLUT4 translocation upon SEC16A depletion was secondary to dysregulation of the exit of proteins from the ER, we determined the effects of knockdown of proteins involved in the formation of COPII vesicles on GLUT4 translocation: SEC13, which interacts with SEC16A on the ER membrane to define sites of COPII coat formation; leucine-rich repeat kinase 2 (LRRK2), which has a role in recruiting SEC16A to the ER membrane; SEC31, which has a role in cargo selection and vesicle formation; and SEC23A and SEC23B, COPII coat proteins (Hughes et al., 2009; Whittle and Schwartz, 2010; Cho et al., 2014). Knockdowns were confirmed by RT-PCR (Fig. S1). Depletion of SEC23B, SEC13, SEC31, or LRRK2 had no effect of insulin-stimulated GLUT4 translocation, suggesting that the blunting of GLUT4 translocation in SEC16A KD adipocytes was not secondary to ER exit site dysregulation (Fig. 3 A). However, knockdown of SEC23A had a significant inhibitory effect on insulin-stimulated GLUT4 translocation (Fig. 3 A). This effect was not secondary to alterations in endosomal membrane trafficking because TR trafficking was not affected by SEC23A knockdown (Fig. 3 B). Thus, only two of the components of the machinery that regulates ER traffic tested, SEC16A and SEC23A, impact GLUT4 trafficking.

To further explore the link between ER exit and GLUT4 translocation, we quantified the effects of knockdowns on secretion of adipsin, an adipocyte-secreted protein that traffics to the PM in vesicles other than GLUT4 (Millar et al., 2000). To synchronize exit from the ER, we incubated cells in Brefeldin A for 45 min to block exit from the ER and collected media after treatment at intervals out to 3 h. We determined the amount of adipsin in the media and associated with cells by quantitative Western blotting (Fig. 3, C and D). Knockdown of SEC16A or SEC13 significantly reduced adipsin secretion, confirming a role for these proteins in ER exit. The fact that knockdowns of SEC16A and SEC13 inhibit secretion to the same extent, but only SEC16A has an effect on insulin-stimulated GLUT4 translocation, establishes that the effect of SEC16A knockdown is not secondary to an effect of ER exit. SEC23A knockdown did not affect adipsin secretion (Fig. 3, C and D; and Fig. S2). Thus, in adipocytes, SEC23A is required for GLUT4 translocation, but not for ER exit.

RAB10 is localized to dynamic tips of reticular ER in proliferative COS-7 cells, functioning in the maintenance of ER dynamics (English and Voeltz, 2013), raising the possibility that RAB10 might have a role in ER exit in differentiated adipocytes. However, RAB10 knockdown, unlike SEC16A knockdown, did not inhibit adipsin secretion, indicating that exit from the ER in adipocytes occurs independent of RAB10 (Fig. 3, C and D). Therefore, the function of SEC16A in transport from the ER is independent of RAB10, suggesting that in differentiated adipocytes there are at least two functional pools of SEC16A: one required at the ER and the other for GLUT4 translocation.

We used ZEISS airy scanning confocal microscopy (lateral resolution of ∼140 nm) to investigate the colocalization of endogenous GLUT4 and endogenous SEC16A. These proteins are located together in the perinuclear region of adipocytes, with no localization in vesicles (puncta) distributed throughout the cytosol (Fig. 4 A). In the perinuclear region, SEC16A was in linear structures abutting GLUT4-containing membranes, and insulin stimulation did not affect the morphology of either the SEC16A or GLUT4 compartments (Fig. 4, A–D). SEC16A labeling defined the “edges” of the GLUT4-containing compartments, although line-scan analyses demonstrated that SEC16A does not colocalize to the same pixels as GLUT4 (Fig. 4, E and F; and Fig. S3).

Ectopically expressed SEC16A-GFP had a distribution identical to that of endogenous SEC16A, encircling the perinuclear accumulation of GLUT4 in basal and insulin-stimulated adipocytes, validating the use of this construct as a reporter for endogenous SEC16A (Fig. 4, G and H).

To determine whether SEC16A-labeled linear structures adjacent to the GLUT4 are SEC16A bound to the surface of ER tubules, we compared the localization of endogenous SEC16A and GLUT4 to the ER labeled by expression of SEC61B-GFP (Fig. 5 and Fig. S4). The linear SEC16A staining in the perinuclear region did not significantly overlap with tubular ER in either basal or insulin-stimulated cells, suggesting these structure are not ER tubules. Furthermore, there was no overlap or obvious relationship between ER structures and the perinuclear tubular accumulation of GLUT4.

Endosomes form contact sites with the ER that are spatially and temporally linked to vesicle fission (Rowland et al., 2014). Thus, the effect of SEC16A knockdown on GLUT4 translocation might be driven by changes in ER morphology (i.e., reduced reticular ER). To address this possibility, we examined the effect of SEC16A knockdown on the morphology of the ER assessed by SEC61B-GFP (Fig. 6, A and B). SEC16A knockdown did not grossly affect SEC61B-GFP morphology in either basal or insulin-stimulated cells. These data establish that the effect of SEC16A knockdown on GLUT4 was not secondary to a pronounced change in ER reticular morphology.

We used an in situ proximity ligation assay (PLA) to probe the proximity of RAB10 and SEC16A in intact adipocytes. This microscopy-based technique detects the proximity of two antibodies within 40 nm of one another (Söderberg et al., 2006). Because the available anti-RAB10 antibodies do not work in immunofluorescence, we cotransfected adipocytes with FLAG-RAB10 and GFP-SEC16A and used anti-FLAG and anti-GFP antibodies for in situ PLA. More than 90% of the transiently transfected cells expressed both GFP-SEC16A and FLAG-RAB10. Thus, we restricted our analysis of the in situ PLA signal to GFP-positive cells. Cells on the same coverslips that did not express GFP were used as controls for nonspecific in situ PLA signal.

The in situ PLA signal in unstimulated adipocytes expressing GFP-SEC16A was twice that measured in cells that did not express GFP, demonstrating proximity of SEC16A and RAB10 in the absence of insulin stimulation (Fig. 7, A–C). The in situ PLA signal in basal adipocytes overlapped with, but was not restricted to, the perinuclear accumulation of SEC16A-GFP. Insulin treatment increased both the overall PLA signal (Fig. 7 B) and the overlap of the in situ PLA signal with the perinuclear SEC16A-GFP signal (Fig. 7 C). These results demonstrate that the proximity of SEC16A and RAB10, predominantly in the perinuclear region, is enhanced by insulin, consistent with SEC16A acting as an effector of RAB10. There was no in situ PLA signal in cells cotransfected with empty GFP vector and FLAG-RAB10, indicating that the in situ PLA signal between GFP-SEC16A and FLAG-RAB10 specifically reflected proximity between the tagged proteins (Fig. 7 D).

Knockdown of AS160, the RAB10 GAP, results in an increase of GLUT4 in the PM of basal adipocytes because of an increase in RAB10-GTP (Sano et al., 2007). The increase in PM GLUT4 in AS160 knockdown adipocytes is rescued by RAB10 knockdown, consistent with RAB10 being the appropriate target of AS160 in adipocytes (Fig. 7 E; Sano et al., 2007). If SEC16A is a RAB10 effector required for GLUT4 translocation, then knockdown of SEC16A will rescue the increased GLUT4 in the PM of AS160 knockdown by functionally mitigating the effects of an increased pool of active RAB10. Knockdown of SEC16A rescued the increase in insulin-independent GLUT4 redistribution to the PM in AS160 knockdown cells, consistent with SEC16A functioning downstream of RAB10 (Fig. 7 E). There was no additive effect of simultaneously knocking down both SEC16A and RAB10 compared with either of the single knockdowns alone (Fig. 7 F), indicating that they function in the same pathway to regulate GLUT4 trafficking. Additionally, there was no compensation for the knockdown of SEC16A via changes in either RAB10 or AS160 at the protein level (Fig. S5).

RAB10 knockdown blunts the insulin stimulated increase in GLUT4 exocytosis (Sano et al., 2007). As expected of the RAB10 effector mediating insulin control of GLUT4 trafficking, knockdown of SEC16A reduced the exocytosis rate of GLUT4 in insulin-stimulated adipocytes to a similar degree as RAB10 knockdown (Fig. 7, G and H). Collectively, these data suggest that SEC16A acts downstream of RAB10 to mediate the effects of insulin on GLUT4 trafficking.

Past work has shown that RAB10 functions to increase GLUT4 in the PM at a step before fusion of the GLUT4-containing vesicles with the PM (Bai et al., 2007; Sano et al., 2007; Sadacca et al., 2013). The prevailing view has been that RAB10 functions to regulate GSV docking to the PM. In TIRF microscopy, imaging proteins within ∼150 nm of the PM, there is little colocalization of GLUT4 and SEC16A in either insulin-stimulated or unstimulated adipocytes, demonstrating that SEC16A is not a component of the GSVs that ferry GLUT4 to the PM (Fig. 8, A and B).

The effect of insulin to increase GLUT4 to the PM is reflected by an increase of GLUT4 within the TIRF zone (Fig. 8 A). Insulin stimulation did not increase SEC16A at the PM when measured as the integrated intensity of anti-SEC16A fluorescence in TIRF microscopy or by the density of SEC16A puncta in the TIRF zone (Fig. 8, C and D). These data indicate that insulin signaling does not recruit SEC16A to the PM.

Because we did not find SEC16A localized with GLUT4 vesicles near the PM, and because we did not find that insulin promotes the localization of SEC16A to the PM, we turned our attention to other potential sites of RAB10-SEC16A regulation of GLUT4 translocation. The known function of SEC16A in regulating COPII-coated vesicle formation suggested that SEC16A might act in the functionally equivalent step of the formation of the GSVs that ferry GLUT4 to the PM. Previous studies have shown that GSVs form from the perinuclear pool of endosomal/TGN GLUT4 (Martin et al., 2000a,b; Zeigerer et al., 2002; Shewan et al., 2003; Karylowski et al., 2004).

To determine if SEC16A knockdown reduced mobilization of GLUT4 from the perinuclear compartment, we used IRAP-TR, reporter of the insulin-regulated GLUT4 trafficking pathway (Subtil et al., 2000). In IRAP-TR, the cytoplasmic domain of IRAP replaces the cytoplasmic domain of human TR in IRAP-TR, conferring GLUT4-like trafficking on the chimera (Subtil et al., 2000; Jordens et al., 2010). The well-developed methods for characterizing the kinetics of TR trafficking can be applied to study IRAP-TR trafficking (Subtil et al., 2000).

Fluorescently labeled transferrin internalized by IRAP-TR accumulates in a perinuclear compartment that overlaps with perinuclear HA-GLUT4-GFP (Fig. 9 A). Apotransferrin bound to IRAP-TR is released from the cells when the IRAP-TR recycles back to the PM and is a measure of the recycling of IRAP-TR to the PM (Subtil et al., 2000). To measure the depletion of transferrin (Tf) specifically from the perinuclear compartment that colocalizes with GLUT4, we created a region of interest that encircled perinuclear GLUT4. The amount of Tf in this region decreased over time as vesicles bud and traffic to the PM (Fig. 9 A). Mobilization of Tf from the perinuclear compartment of SEC16A knockdown cells was decreased compared with control cells, consistent with SEC16A having a role in GLUT4 mobilization from this perinuclear region (Fig. 9 B). RAB10 knockdown had a similar effect on insulin-stimulated mobilization of IRAP-TR from the perinuclear region (Fig. 9 B).

SEC16A encircles GLUT4 in the perinuclear region of adipocytes (Fig. 4). We next determined if SEC16A copurifies with GLUT4-containing membranes. We isolated GLUT4-containing membranes from both basal and insulin-stimulated adipocytes stably expressing HA-GLUT4-GFP by detergent-free anti-GFP magnetic microbead immunoabsorption. We used two different concentrations of beads to achieve either near-complete or a 50% immunoabsorption of GLUT4. GLUT4 is in equilibrium among several different intracellular compartments (Karylowski et al., 2004). All of these compartments will be absorbed with near-complete GLUT4 immunoabsorption (high amount of beads). We reasoned that submaximal immunoabsorption would identify proteins that are specifically enriched in membranes containing the highest concentration of GLUT4, such as the perinuclear compartment. When we reduced the immunoabsorption of GLUT4 to ∼50% in basal adipocytes, the co-absorption of AS160 and IRAP, two proteins known to traffic with GLUT4, were also reduced by ∼50% (Fig. 9, C and D). The relative co-absorption of calnexin and syntaxin 6 with GLUT4 was decreased by 90% with reduced immunoabsorption of GLUT4, suggesting that these proteins, markers of the ER and a subdomain of the TGN, respectively, are not as enriched in GLUT4-containing membranes as AS160 and IRAP (Fig. 9, C and D). The observation that TGN46, another TGN marker, does not co-absorb with GLUT4-containing membranes demonstrates that GLUT4 is concentrated in a selective domain of the TGN. Interestingly, the amount of SEC16A coimmunoabsorbed with GLUT4 was not reduced, suggesting SEC16A is highly enriched with the subset of GLUT4-containing membranes that are efficiently immunoabsorbed from basal adipocytes (Fig. 9, C and D).

The results were similar for insulin-stimulated adipocytes, except for SEC16A, where there was a significant decrease in the amount of co-absorbed SEC16A in conditions of reduced GLUT4 immunoabsorption (Fig. 9, C and E). Upon insulin stimulation, GLUT4 becomes more concentrated in GSVs. SEC16A is not a component of GSVs (Jedrychowski et al., 2010), and this decrease in SEC16A association with GLUT4 compartments in insulin-stimulated cells may reflect this redistribution of GLUT4.

It is of note that there is significant SEC16A that does not co-absorb with GLUT4 (in the flow through) under conditions when >90% of GLUT4 is immunoabsorbed (Fig. 9 C). Two biochemical pools of SEC16A were expected because of the independent roles of SEC16A in regulating ER-to-Golgi trafficking, and GLUT4 mobilization from the perinuclear recycling endosome/TGN would require it to reside in these two distinct compartments.

If SEC16A is involved in GSV formation, then SEC16A knockdown would change the composition of GLUT4-containing vesicles. To determine if this was the case, we isolated GLUT4-containing membranes from adipocytes stably expressing HA-GLUT4-GFP by detergent-free immunoabsorption with anti-GFP magnetic microbeads (Fig. 9 E). SEC16A knockdown significantly reduced the association of AS160 and RAB10 with the GLUT4-containing compartments in insulin-stimulated cells, without an effect in the basal condition. These data provide further evidence for the role of SEC16A in AS160-RAB10 regulation of GLUT4 trafficking. Coimmunoabsorption of IRAP was unchanged by SEC16A knockdown. This was expected, because IRAP traffics by the same pathway as GLUT4, and perturbations affecting one are known to affect the other (Eguez et al., 2005; Sano et al., 2007; Jordens et al., 2010). However, TR, a marker of general endocytic trafficking, was unchanged by SEC16A knockdown in either basal or insulin-stimulated conditions, establishing that SEC16A knockdown did not result in a significant shift of GLUT4 from specialized membrane compartments to the more general Tf-positive endosomes (Fig. 9 F).

Insulin triggers transport of blood glucose into adipocytes and muscle cells by stimulating a redistribution of GLUT4 from intracellular compartments to the PM. The amount of GLUT4 in the PM is limiting for glucose uptake, and therefore the postprandial lowering of blood glucose by disposal of glucose into adipose and muscle requires a persistent elevation of PM GLUT4 that parallels increased insulin levels. Because GLUT4 continually cycles between the PM and intracellular compartments, the maintenance of elevated cell-surface GLUT4 in the presence of insulin requires accelerated biogenesis of the specialized vesicles that transport GLUT4 to the PM. Identifying the protein machinery that regulates GLUT4 trafficking will provide insights into the molecular mechanism underlying the control of glucose metabolism and thereby provide a framework for understanding its disruption in insulin-resistance.

Here, we establish SEC16A as a RAB10 effector involved in GLUT4 translocation. Our data support a model in which RAB10, activated downstream of insulin stimulation, interacts with SEC16A to promote the formation of the GSVs that ferry GLUT4 to the PM of adipocytes (Fig. 10). The RAB10-SEC16A module, by stimulating GSV formation, supports a persistent elevated level of GLUT4 in the PM during continued insulin stimulation. Insulin stimulation of GSV formation is impaired by knockdown of RAB10, SEC16A, or the COPII coat subunit SEC23A, resulting in a blunting of the steady-state redistribution of GLUT4 to the PM of stimulated adipocytes. Disruption of the RAB10-SEC16A module by knockdown of the individual proteins inhibits GLUT4 translocation by ∼50%. This partial effect is consistent with a model in which insulin controls the amount of GLUT4 in the PM by regulating multiple trafficking steps that likely include, in addition to the stimulated formation of the GSVs via RAB10-SEC16A function, the docking and fusion of the GSVs to the PM. Recent studies support insulin control of exocyst function that contributes to GLUT4 redistribution by regulating the docking of GLUT4 vesicles to the PM (Karunanithi et al., 2014; Sano et al., 2015).

We have used several different approaches to test the hypothesis that SEC16A is an effector of RAB10 in the regulation of GLUT4 trafficking in adipocytes. First, SEC16A knockdown recapitulates the decrease in GLUT4 exocytosis seen with RAB10 knockdown, and SEC16A knockdown combined with RAB10 knockdown had no additional effect on GLUT4 translocation compared with either single knockdown. Second, both SEC16A knockdown and RAB10 knockdown restore GLUT4 retention in unstimulated AS160 knockdown adipocytes. Knockdown of AS160, a negative regulator of RAB10, results in an increase of GLUT4 in the PM in the absence of insulin stimulation, and this increase is reversed by RAB10 knockdown. SEC16A also reverses the effect of AS160 knockdown on GLUT4, which strongly supports the conclusion that SEC16A functions downstream of active RAB10. Third, similar to RAB10 knockdown, SEC16A knockdown does not affect TR trafficking or the insulin-signal transduction cascade.

It has previously been shown in TIRF microscopy studies that one effect of insulin is to promote association (tethering/docking) of GSVs with the PM (Lizunov et al., 2013) and that the AS160-RAB10 signaling module acts at a prefusion step of GLUT4 trafficking, presumably regulating the docking/tethering of GSVs to the PM (Gonzalez and McGraw, 2006; Bai et al., 2007; Sano et al., 2007; Sadacca et al., 2013). Therefore, we expected that SEC16A, as a RAB10 effector, would function at the PM. However, our data support a model in which SEC16A acts at the perinuclear recycling endosome/TGN to stimulate the formation of GSVs, thereby mobilizing insulin-regulated cargo for delivery to the PM. SEC16A is localized immediately adjacent to GLUT4 in the perinuclear compartment. There is no significant colocalization between GLUT4 and SEC16A in the TIRF zone (within 150 nm of the PM) or with GLUT4-containing vesicles distributed throughout the cytosol, demonstrating that SEC16A is not associated with GSVs, as might be anticipated if SEC16A regulated the association of these vesicles with the PM. Furthermore, insulin does not stimulate a redistribution of SEC16A to the TIRF zone, as might be expected if SEC16A regulated GSV docking in an insulin-sensitive manner. Collectively, these data are not consistent with SEC16A functioning to regulate GSV docking to the PM.

SEC16A has a role in transport from the ER by regulating the assembly of COPII vesicle components on the ER membrane (Connerly et al., 2005; Sprangers and Rabouille, 2015). Our data support RAB10-SEC16A functioning in the stimulated mobilization of GLUT4 from a perinuclear recycling endosome/TGN compartment. There are previous studies showing that insulin regulates the formation of GLUT4 vesicles, although this has not been linked to RAB10 or SEC16A (Huang et al., 2013). Our data support SEC16A acting in the biogenesis of GSVs. First, SEC16A is localized adjacent to GLUT4 in the perinuclear recycling endosome/TGN compartment, the site of GSV formation. Second, insulin increases the proximity of SEC16A and RAB10 in the perinuclear compartment (detected by in situ PLA), consistent with SEC16A acting at the perinuclear compartment of insulin-stimulated cells. Third, SEC16A or RAB10 knockdown slow the insulin-stimulated exit of IRAP-TR from the perinuclear compartment, providing direct support for RAB10-SEC16A functioning in the mobilization of cargo of the insulin-regulated pathway from the perinuclear compartment. Finally, SEC16A knockdown decreased the association of AS160 and RAB10 with GLUT4-containing membranes in insulin-stimulated adipocytes, providing support for reduced GSV formation, because both AS160 and RAB10 are components of GSVs (Jedrychowski et al., 2010). The observations that in unstimulated adipocytes neither GLUT4 trafficking nor the association of RAB10 and AS160 with GLUT4 membranes is affected by SEC16A knockdown demonstrate that SEC16A is not required for formation of GSVs in basal conditions when demand is low but is required only when the demand for GSVs is high, such as in the context of insulin stimulation. Our results do not preclude RAB10 from also acting at other prefusion trafficking steps along the GLUT4 itinerary. It has recently been reported that SEC15b, an exocyst component that interacts with RAB10 in yeast two-hybrid analyses, is also an effector of RAB10; however, the function of RAB10-SEC15b in GLUT4 trafficking has not been extensively characterized (Sano et al., 2015).

Our results establish that in differentiated adipocytes, the role of SEC16A in the ER is independent of its role in GLUT4 trafficking. Knockdown of SEC13, which functions with SEC16A to regulate exit from the ER, does not affect GLUT4 trafficking but causes the expected defect in transport from the ER. In addition, knockdown of LRRK2, a kinase that recruits SEC16A to the ER, does not inhibit GLUT4 translocation. These results uncouple general inhibition of ER exit from insulin regulation of GLUT4 trafficking. Finding SEC16A functions in both exit from the ER and GLUT4 translocation suggests there must be two functional pools of SEC16A. In agreement with this proposal, we show biochemically two pools of SEC16A in adipocytes, one that localizes to GLUT4 compartments and one that does not.

Transient knockdown of other COPII components, SEC23B and SEC31, did not blunt insulin-stimulated translocation. However, knockdown of SEC23A, another COPII component, blunted insulin-stimulated GLUT4 translocation. Finding a subset of the COPII vesicle coat machinery components involved in GLUT4 translocation is consistent with there being functional subcomplexes of COPII components. Functional subcomplexes of the clathrin adaptor proteins and the retromer provide precedents for the possibility of functional COPII subcomplexes. In this regard, the number of paralogs of COPII components in mammalian cells makes possible a large variety of COPII coats to fulfill cargo and or tissue-specific functions (Khoriaty et al., 2012). For example, mutations in SEC23B cause congenital dyserythropoietic anemia type II, suggesting a specific role for SEC23B in erythroblasts of the bone marrow (Khoriaty et al., 2012).

SEC16A association with GLUT4 compartments is reduced by insulin stimulation, seemingly inconsistent with its proposed role in the formation of GLUT4-containing vesicles. However, during insulin stimulation, GLUT4 is mobilized from intracellular storage compartments to GSVs and SEC16A is not a component of GSVs (Jedrychowski et al., 2010), thereby explaining the decrease in SEC16A association with GLUT4-containing membranes. These results are in agreement with SEC16A function at the ER, where it has a role in generating COPII vesicles without being a component of these vesicles (Budnik and Stephens, 2009; D’Arcangelo et al., 2013).

The ER has been proposed to serve as a general platform for membrane fission and fusion events (Rowland et al., 2014). Given the role of SEC16A in ER transport, it is conceivable that the effects of SEC16A knockdown are caused by changes in ER morphology that disrupt the dynamics of membranes containing GLUT4. However, ER morphology is not altered in SEC16A knockdown cells. Furthermore, it is highly unlikely that disruption of an ER function required for general membrane fission and fusion would specifically impair GLUT4 trafficking but have no effect on general vesicular trafficking. However, SEC16A knockdown specifically inhibits GLUT4 translocation but has no effect on TR.

In addition to the role of RAB10 as a key mediator of GLUT4 trafficking in adipocytes, RAB10 functions in other cell types in the regulation of other membrane trafficking processes: sorting of recycling cargo in polarized cells (Babbey et al., 2006, 2010; Lerner et al., 2013; Deen et al., 2014; Xu et al., 2014), polarized membrane trafficking required for axon development (Habraken et al., 1999), TLR4 trafficking in macrophages (Kasprzak et al., 1998), and phagosome maturation (Penn et al., 1990). Although the specific vesicular cargo and molecular machinery required for each of these functions are quite disparate, what they have in common is that they are all regulated trafficking events in which RAB10 regulates vesicle delivery to the PM.

Recently, it was found that SEC16A functions as an integrator of growth factor signaling and secretion in HeLa cells, acting as the central node in a feed forward loop and allowing growth factors to prepare cells for the increased secretory cargo load that accompanies proliferation (Tillmann et al., 2015). Even though our data demonstrate that the effect of SEC16A knockdown on GLUT4 trafficking in adipocytes is not secondary to perturbations of flux through the ER, it is provocative to consider how SEC16A’s role as a potential integrator of hormonal cues in proliferative cells has been adapted for use in terminally differentiated adipocytes, in which proliferation is not the end goal of growth factor signaling.

3T3-L1 fibroblasts were maintained in culture and differentiated into adipocytes as previously described (Zeigerer et al., 2002). For some experiments, previously described 3T3-L1 cell lines stably expressing shRNA sequences directed against RAB10 or AS160 were used (Eguez et al., 2005; Sano et al., 2007). Immunoabsorption experiments were performed using 3T3-L1 cell lines stably expressing HA-GLUT4-GFP.

The HA-GLUT4-GFP, TR, IRAP-TR, FLAG-RAB10, SEC61B-GFP, and SEC16A-GFP DNA plasmids have been described previously (Lampson et al., 2001; Bhattacharyya and Glick, 2007; Friedman et al., 2010; Jordens et al., 2010; Giordano et al., 2013; Sadacca et al., 2013). SEC16A plasmids contained sequences for human SEC16A and were resistant to knockdown by mouse SEC16A-directed siRNA sequences. For rescue experiments, we generated SEC16A-GFP in which we mutated GFP to ablate the fluorescence by the following mutations in the enhanced GFP of the SEC16A-GFP construct: T66A, Y67A, and G68A. This was done using the QuikChange II Site-Directed Mutagenesis kit (Agilent Technologies). The siRNA constructs were, wherever possible, as previously published.

Sequences were as follows: RAB10: si251, 5′-GCAUCAUGCUAGUGUAUGA-3′ (same sequence as shRNA expressed by RAB10 KD cells; Sano et al., 2007); AS160: 5′-GACTTAACTCATCCAACGA-3′ (same sequence as shRNA expressed by AS160 knockdown cells; Eguez et al., 2005); SEC16A: si1, 5′-CTTCAGAATATCAGCTCCCTGGGGCTC-3′, si3, 5′-AGCTGGACTTGCTGGTGGCTGGGCCAA-3′; ARHGEF7: 5′-AGCAGCUUUUUCAUGUCUUGGGGCUC-3′; IQGAP1: 5′-GAGAGGAACUUGUCUGUCACGGCCCUC-3′; GIT1: 5′-GGUGGGUUAGUGUCUGGAGUGGCCCAG-3′; PFN1: 5′-AAAAGCCAAAGGGAACGGUGGGGGGAA-3′; ANXA2: 5′-AGGCCCAAAAUCACCGUCUCCAGGUGG-3′; SEC13: 5′-GGAGGCUGACACAGAACCCUGGCCCUU-3′; LRRK2: 5′-CAAGAAACUCUCAGGUCGAUUGUCUAA-3′; SEC23A: 5′-AGGUUGUCAUGGUAGCACUUGAUUCUU-3′; SEC23B: 5′-UCUGAAAGACGUAGGGUCUUCUUUGUU-3′; and SEC31: 5′-AUAUAAUACCUCUCCUGUGUUAGGGUU-3′.

When only one siRNA sequence was used, 2 nmol of that siRNA was electroporated. When two siRNAs, 2 nmol of each siRNA was electroporated. Two siRNAs targeting SEC16A were required to achieve this degree of knockdown, whereas only one was required for the other proteins. Measurement of knockdown was performed by quantitative PCR. 3T3-L1 adipocytes were electroporated with siRNAs as described. At 48 h after electroporation, cells were harvested and RNA was extracted using the RNeasy kit from QIAGEN. From extracted RNA, cDNA was made using the RNA to cDNA EcoDry Premix from Takara Bio Inc., and quantitative PCR was performed using a qSTAR primer pair obtained from OriGene.

Indicated cell lines were electroporated with 45 µg of each indicated plasmid and/or 2–4 nmol siRNA as previously described (Zeigerer et al., 2002). Assays were performed 1–2 d after electroporation, as needed to achieve knockdown of the targeted protein as validated by RT-PCR. On the day of the assay, GLUT4 or TR translocation assays were performed as described previously (Lampson et al., 2001; Zeigerer et al., 2002; Karylowski et al., 2004; Sano et al., 2007; Jordens et al., 2010). In brief, cells were washed and incubated in media without sera for 2 h. Cells were then stimulated with 1 nM insulin for 30 min to achieve steady-state surface GLUT4 levels. Cells were fixed with 3.7% formaldehyde for 7–10 min, and an anti-HA antibody was used, without permeabilization, to label HA-GLUT4-GFP on the cell surface. HA staining was visualized with fluorescently tagged secondary antibodies, and total HA-GLUT4-GFP was visualized by direct fluorescence, as described later (Lampson et al., 2000; Karylowski et al., 2004). Mouse anti-HA was purchased from Covance, and secondary antibodies were obtained from Jackson ImmunoResearch Laboratories and Invitrogen.

For experiments where the effect of SEC16A KD on GLUT4 translocation and basal TR levels was binned into quartiles based on residual SEC16A expression, anti-HA primary and secondary staining was followed by permeabilization with 0.25 mg/ml saponin (Sigma-Aldrich) and indirect immunofluorescence with anti-mouse SEC16A antibody (MKA0310; Cosmo Bio Co.) to label endogenous SEC16A. SEC16A staining was visualized with a Cy5-labeled secondary antibody (Thermo Fisher Scientific).

For experiments assessing colocalization of endogenous SEC16A and SEC16A-GFP with endogenous GLUT4, cells were electroporated with SEC16A-GFP DNA as described, followed by permeabilization with 0.25 mg/ml saponin (Sigma-Aldrich) and indirect immunofluorescence with mouse anti-GLUT4 antibody (2213S; Cell Signaling Technology) or rabbit anti-SEC16A antibody (MKA0310; Cosmo Bio Co.) that was visualized with fluorescently tagged secondary antibody.

All epifluorescence images were collected on an inverted microscope at room temperature using a 20× air objective (Leica Biosystems) and a cooled charge-coupled device 12-bit camera. Exposure times (Jordens et al., 2010) and image quantification (Lampson et al., 2001) were performed using MetaMorph image processing software (Universal Imaging), as previously described. GFP and Cy3 fluorescence signals were background corrected and the surface(Cy3):total(GFP) (S:T) GLUT4, TR, or VSVG ratio was calculated for each cell. The S:T values were normalized within each assay to the mean S:T value for the indicated condition to allow for averaging results across multiple biological repeat assays. For experiments where SEC16A siRNA-treated cells were binned into quartiles, the integrated anti-SEC16A (Cy5) fluorescence was measured for each cell, and cells were subsequently binned into quartiles based on residual SEC16A expression. The corresponding mean GLUT4 or TR S:T value for cells in each quartile was calculated and normalized to the wild-type population mean S:T. The levels of residual SEC16A expression in each quartile was also calculated by normalizing the mean anti-SEC16A (Cy5) fluorescence of cells in each quartile to the wild-type population mean anti-SEC16A (Cy5) fluorescence. The residual SE paired Student’s t tests were performed on raw (non-normalized) S:T mean values from multiple assays. One-tailed Student’s t tests were only used when a reasonable expectation could be formed from previous results as to the direction of the measured effect. Two-way analysis of variance was used only where indicated.

Airyscan confocal images of endogenous GLUT4 and SEC16A colocalization were collected on a laser scanning microscope (LSM880; ZEISS) with Airyscan using a 63× objective. All other confocal images were collected on a laser scanning confocal microscope using a 63×, 1.4 NA Plan-Apochromat objective.

Western blot experiments were performed using standard protocols. Cellular protein extracts were collected from 3T3-L1 differentiated adipocytes expressing indicated siRNA or DNA constructs. Antibodies used were as follows: RAB10 (4262S; Cell Signaling Technology); AS160 (07–741; EMD Millipore); ACTIN (AAN01-A; Cytoskeleton, Inc.); AKT (9272S; Cell Signaling Technology); P-AKT (9271S; Cell Signaling Technology); P-AS160 (8881S; Cell Signaling Technology); TR (ab84036; Abcam); IRAP (6918P; Cell Signaling Technology); GFP (Roche); and SEC16A (MKA0310; Cosmo Bio Co.).

Rab10 was cloned using BamHI and XhoI into pGEX-6P-1 plasmid (Promega) with a C-terminal hexa-his tag, GST-Rab10-His. GST-tagged Rab10-His protein was expressed in Escherichia coli BL21 (DE3) cells (Invitrogen) during overnight induction at 18°C. Bacterial pellets were resuspended in 20 mM Hepes, pH 7.2, 100 mM KCl, 1 mM TCEP, and 0.1% IGEPAL supplemented with an antiprotease cocktail (Roche) and 1 mM phenylmethylsulfonyl fluoride. The resuspended cell pellets were lysed on ice by sonication followed by ultracentrifugation at 100,000 g for 1 h. Fusion protein was purified by GST affinity chromatography. The N-terminal GST tag was removed by digestion with PreScission protease (Invitrogen), and Rab10-His was separated from GST and uncleaved fusion protein on glutathione-agarose. After overnight dialysis in 20 mM Hepes, pH 7.2, 100 mM KCl, and 1 mM DTT, Rab10-His was further purified using a size exclusion chromatography using Superdex200 column (GE Healthcare) in 20 mM Hepes, pH 7.2, 100 mM KCl, and 1 mM DTT. Purified protein was stored at −80°C.

Purified Rab10-His was loaded with GTPγS by incubation with nucleotide exchange buffer (20 mM Hepes, pH 7.5, 100 mM KCl, 5 mM MgCl₂, and 1 mM DTT) containing 1 mM GTPγS for 1 h at room temperature under constant agitation. After loading, GTPγS-Rab10-His was incubated with equilibrated Ni-NTA beads at 25°C for 1 h under constant agitation. After incubation, GTPγS-Rab10-His bound to beads was washed three times in wash buffer (20 mM Hepes, pH 7.5, 100 mM KCl, 20 mM imidazole, and 1 mM DTT).

3T3-L1 adipocytes were incubated with serum-free media for 2 h and then stimulated with 1 nM insulin for 30 min. Cells were lysed in RIPA buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 0.5% sodium deoxycholate, 1% NP-40, and 0.1% SDS). GTPγS-Rab10-His beads were incubated with insulin-stimulated 3T3-L1 adipocyte cytosol at 4°C for 1 h under constant agitation. After incubation with cytosol, the GTPγS-Rab10-His beads and cytosol were passed through a column. The GTPγS-Rab10-His column was washed in 10 column volumes of wash buffer (20 mM Hepes, pH 7.5, 100 mM KCl, 20 mM imidazole, and 1 mM DTT) to remove nonspecific binding. GTPγS-Rab10-His and interacting proteins were eluted using three column volumes of elution buffer (20 mM Hepes, pH 7.5, 100 mM KCl, 0.5 M imidazole, and 1 mM DTT). GTPγS-Rab10-His fractions were examined by SDS-PAGE and subsequently sent for mass spectrometry.

3T3-L1 adipocytes were incubated at 37°C with 5 µg/ml Brefeldin A for 45 min to block ER exit. Cells were washed and reincubated in serum-free media without Brefeldin. The media (secreted proteins) was collected and cells solubilized after 45, 60, 120, and 180 min of recovery. The amount of adipsin in the media and associated with cells was determined by quantitative Western blotting. To calculate the rate of secretion of adipsin from cells, the amount in the media at each time point was normalized to the cell associated adipsin at the start of the incubation in Brefeldin-free media.

3T3-L1 adipocytes were electroporated with 45 µg of each indicated plasmid. Cells were incubated in serum free media for 2 h and then stimulated with 1 nM insulin for 30 min where indicated. Cells were washed three times with PBS before fixation for 7–10 min in 3.7% formaldehyde. Cells were permeabilized with 0.25 mg/ml saponin and immunostained with mouse anti-FLAG antibody (F7425; Abcam) and rabbit anti-GFP antibody (ab6556; Abcam) for 30 min, after which PLA was performed according to the instructions of the manufacturer (Duolink In Situ Orange Starter kit Mouse/Rabbit; Sigma-Aldrich). For quantification, signals in at least 30 cell images were quantified in each experiment, for three experiments. Images are representative of the results of three independent experiments.

3T3-L1 adipocytes electroporated with HA-GLUT4-GFP and indicated siRNA were serum starved for 2 h and incubated with 1 nM insulin for 1 h. Cells were then incubated with anti-HA antibodies (100 µg/ml) in the presence of 1nM insulin for 5, 10, 15, 30, 45, and 60 min. Cells were fixed with 3.7% formaldehyde for 7–10 min, permeabilized with 0.25 mg/ml saponin, and then stained with anti–mouse Cy3 to label the entire pool of antibody-labeled HA-GLUT4-GFP. Cells were imaged and the Cy3-to-GFP (total GLUT4) ratio was determined over time. The rate of exocytosis was determined by measuring the slope of the curve.

For TIRF microscopy experiments, dual-color TIRF and epifluorescence images were acquired on a Nikon TiE inverted microscope with perfect focus mechanism using a 63×, 1.49 NA objective and Neo sCMOS camera (Andor Technology). Excitation wavelengths at 488 for GFP and 561 for Cy3 were used.

3T3-L1 adipocytes were electroporated with 45 µg of HA-GLUT4-GFP. Cells were incubated in serum-free media for 2 h and then stimulated with 1 nM insulin for 30 min where indicated. Cells were fixed for 7–10 min in 3.7% formaldehyde, permeabilized with 0.25 mg/ml saponin, and immunostained with anti–mouse SEC16A antibody (MKA0310; Cosmo Bio Co.). SEC16A staining was visualized with a Cy3-tagged secondary antibody.

The Pearson’s correlation coefficient (r) for HA-GLUT4-GFP and SEC16A puncta in the TIRF zone was calculated using MetaMorph image processing software by generating binary masks of GLUT4 (GFP) TIRF puncta above a fixed threshold and SEC16A (Cy3) TIRF puncta above a fixed threshold.

To calculate the SEC16A TIRF/epifluorescence ratio, the integrated anti-SEC16A (Cy3) fluorescence signal in TIRF was calculated, background corrected, and normalized to the total anti-SEC16A (Cy3) signal in epifluorescence that was also background corrected. ImageJ software was used for image quantification.

For quantification of the number of SEC16A-postive puncta in TIRF microscopy, binary masks of SEC16A TIRF puncta above a fixed threshold were generated. A size threshold was applied to only include puncta with an area >2 × 2 pixels in the analysis, and the number of puncta were counted using ImageJ software.

3T3-L1 adipocytes electroporated with IRAP-TR, and indicated siRNA were treated overnight with an antibody against mouse TR to block endogenous TR expression. The next day, cells were incubated in serum free DMEM containing 1 nM insulin and 5 µg/ml Tf-Alexa⁴⁸⁸ or Tf-Cy3 for 1 h to label the entire pool of IRAP-TR. They were then washed three times in PBS and incubated in serum-free DMEM containing 1 nM insulin and 8 µg/ml unlabeled Tf for indicated time points. Cells were fixed with 3.7% formaldehyde for 7–10 min, permeabilized with 0.25 mg/ml saponin, and then stained with an antibody against human to label the total IRAP-TR pool. Staining was visualized with fluorescently tagged secondary antibodies. A mask was generated by creating a circle of fixed size that tightly encircled both the nucleus and those pixels that colocalized with perinuclear HA-GLUT4-GFP. This circle was drawn around the nucleus of IRAP-TR–expressing cells such that it encircled both the nucleus and perinuclear regions, allowing us to restrict our analysis to this region. The total Tf-Cy3/Alexa⁴⁸⁸ signal within this restricted region was measured and ratioed to the amount of total IRAP-TR expression in this region. The decrease in this ratio over time is indicative of IRAP-TR mobilization from this perinuclear region. Each siRNA treatment was normalized to its starting ratio to control for differences in Tf-Cy3/Alexa⁴⁸⁸ uptake. These ratios were then plotted, and the data were fit to an exponential decay curve.

3T3-L1 adipocytes that stably express HA-GLUT4-GFP were electroporated with or without 4 nmol SEC16A siRNA, as indicated. Cells were washed and incubated in media without sera for 2 h. Cells were then stimulated with or without 1 nM insulin for 30 min to achieve steady-state surface GLUT4 levels. Cells were lysed in cold HES buffer (20 mM Hepes, 1 mM EDTA, 250 mM sucrose, and protease inhibitors) using a cell scraper on ice. They were homogenized by subsequent passaging through 23G and 27G 1 1/4 syringes on ice. Lysates were spun at 700 g for 10 min at 4°C. Supernatant was collected and incubated for 30 min at 4°C with µMACS anti-GFP MicroBeads (Miltenyi Biotec). In different experiments as noted, different amounts of microbeads were used to vary the efficiency of immunoabsorption of GLUT4-containing membranes. The sample was loaded onto a MACS column (Miltenyi Biotec) and placed in the magnetic field of a µMACS Separator in which the magnetically labeled HA-GLUT4-GFP–containing compartments were retained. Columns were washed with PBS and eluted with denaturing elution buffer (50 mM Tris HCl, pH 6.08, 50 mM DTT, 1% SDS, 1 nM EDTA, 0.005% bromphenol blue, and 10% glycerol). Western blots were performed on lysates, flow through, and elution fractions, probing for the indicated proteins. Band intensity was quantified using ImageJ and normalized to the amount of GFP eluted in each sample to control for immunoabsorption efficiency. Amounts of indicated proteins in the elution fraction were normalized to the amount of that protein in the control sample.

Fig. S1 shows RT-PCR quantification of SEC16A, IQGAP1, ARHGEF7, GIT1, PFN1, ANXA2, SEC13, LRRK2, SEC23A, SEC23B, and SEC31 knockdown. Fig. S2 shows adipsin secretion is blunted by SEC16A and SEC13 knockdown, but not by SEC23A or RAB10 knockdown. Fig. S3 shows SEC16A encircles perinuclear GLUT4 in adipocytes. Fig. S4 shows the ER marker Sec61B-GFP does not colocalize with perinuclear SEC16A or encircle perinuclear GLUT4. Fig. S5 shows SEC16A knockdown does not alter RAB10 or AS160 expression. Additional data are available in the JCB DataViewer at http://dx.doi.org/10.1083/jcb.201509052.dv.

Mass spectrometry was performed at the Cornell University Proteomics and Mass Spectrometry Core, TIRF microscopy was performed in the Rockefeller University Microscopy Core, and confocal microscopy was performed in the Weill Cornell Medicine Optical Microscopy Core Facility. We thank Timothy Ryan, Jeremy Dittman, Fred Maxfield, Eva Gonzalez, Reema Vazirani, Nazish Abdullah, Muheeb Beg, Jennifer Wen, and Christopher Torsitano for helpful discussions and critically reading the manuscript.

This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (grant RO1 DK52852 to T.E. McGraw). J. Bruno was supported by a Medical Scientist Training Program grant from the National Institute of General Medical Sciences of the National Institutes of Health and grant T32GM007739 to the Weill Cornell/Rockefeller/Sloan Kettering Tri-Institutional MD-PhD Program, and A. Brumfield is supported by National Institutes of Health training grant 5T32GM008539-19.

The authors declare no competing financial interests.

Author contributions: J. Bruno, A. Brumdield, and N. Chaudhary designed and performed experiments and contributed to the writing of the manuscript. D. Iaea performed RAB10 purification and affinity chromatography. T.E. McGraw conceived of the project, designed experiments, supervised the project, and wrote the manuscript.